Kisung Nam

Kisung Nam

From genetic variation to biological discovery.

I develop statistical methods to understand how genetic variation shapes human disease. Building on my work in statistical genetics and machine learning, I aim to use AI to uncover biological mechanisms and improve genetic risk prediction.

Explore publications Download CVMore about me

Research approach

Statistics and AI, grounded in biology.

Statistical genetics

Methods for rare variants, large biobanks, and genetic risk prediction.

RareEffectMeta-SAIGEKorean Biobank analysis

Machine learning & AI

Statistical approaches to understanding learning, optimization, and model performance.

Predicting human-gain curves

Future direction AI for genomics

Biological discovery

Connecting genetic signals to disease mechanisms and more informative risk prediction.

Discuss a collaboration →

Selected Publications

All publications →
Predicting the human-gain curve of a proxyRepeated human ratings of each item settle into a quality estimate. Two forecasts of human gain are drawn before optimization. As items are reweighted toward higher proxy scores, the measured held-out gains land on the highlighted two-coefficient forecast as it levels off, while a first-order forecast keeps climbing and misses by a widening gap.human ratinghuman gainproxy scoreoptimization (KL)

2026 Machine learning

Predicting human-gain curves

Using repeated human ratings to predict how human-rated performance changes under local proxy optimization.

NeurIPSSpotlight (Top 1.3%)To appear
From rare variant effects to high-risk individualsUnstable single-variant estimates shrink into a gene-level range while keeping their positive or negative direction. Scores built from these rare variant effects move some carriers past the high-risk threshold of a polygenic risk distribution.variant effectspolygenic riskhigh risk

2026 Statistical genetics

RareEffect

Using gene-level information to estimate rare variant effects, improve genetic risk prediction, and identify high-risk individuals.

Nature GeneticsPaper
Shared genetics and causal links between oral and systemic diseaseAn arch of 21 orofacial diseases faces a column of systemic diseases. Undirected roots from shared genetic loci reach dental caries, periodontitis, and systemic diseases alike, while directed arrows run from Sjögren's syndrome to lung cancer and stroke.orofacialsystemicshared geneticscausal

2026 Genomic discovery

Orofacial disease analysis

Connecting oral and systemic health through the shared genetic architecture of 21 orofacial diseases.

Nature CommunicationsPaper
Combining calibrated rare variant evidence across biobanksGene-based association signals in UK Biobank and All of Us each stay below the significance threshold. Combined by meta-analysis, two genes rise above it; a calibration sweep pulls two spurious signals from case-control imbalance back below the threshold while the real associations remain.significancecalibrationUK BiobankAll of Usmeta-analysis

2025 Statistical genetics

Meta-SAIGE

Combining rare variant evidence across biobanks while controlling false positives for low-prevalence diseases.

Nature GeneticsPaperCodeResults
A pleiotropy skyline across 76 traits in the Korean biobankFor each of 76 traits in 72,298 Korean individuals, significant associations drop onto their genomic positions and stack. The stacks form a skyline whose tallest tower is the ALDH2 region on chromosome 12, associated with about 30 traits; novel associations are highlighted.chr 12276 traitsALDH2novel loci

2022 Genomic discovery

Korean Biobank analysis

Mapping genetic associations for 76 traits in 72,298 Korean participants and expanding genomic resources for East Asian populations.

Cell GenomicsPaperResults